Context Cells

Contacts

Alberto Calatroni, Dr. Daniel Roggen
Related project page: OPPORTUNITY

Publications

[1] Context Cells: Towards Lifelong Learning in Activity Recognition Systems
Alberto Calatroni, Claudia Villalonga, Daniel Roggen and Gerhard Tröster
in: Proceedings of the 4th European Conference on Smart Sensing and Context (EuroSSC), Springer, 2009

General description

Context Cells have been developed within the OPPORTUNITY project to demonstrate principles of opportunistic activity recognition. Context Cells are able to learn how to recognize activities online and feature machine learning algorithms as a means to infer gestures/activities.

The nodes feature a main board hosting accelerometer (ADXL335), a magnetometer (in the latest version) and a RISC Jennic CPU (JN5139/5148) with integrated ZigBee transceiver. Furthermore, the boards can be connected to a miniaturized external IMU which allows to get the real time orientation (for details on the IMU see [2]).

Hardware Components

1. Jennic Microcontroller+Transceiver
2. External (optional) IMU [2]
   
3. FTDI232 chip for USB connection
4. Reed switch for power on/power off
5. Program/Reset buttons
6. Micro-USB plug
7. Buzzer
8. MicroSD card slot
9. SRAM (256kbit)
   
10. External crystal for time reference and synchronization
11. Accelerometer (ADXL 335)
12. LTC3455 power controller/regulator/battery charger and monitor

Highlights

• custom firmware allows for time synchronization of many boards with coordinator
• usable as datalogger (or synchronous dataloggers)
  
• easy communication with PC through virtual COM port
• multifunctional USB connection (recharges the battery, allows firmware reprogramming and communication with the board)
• SRAM for extra data buffering
• firmware primitives for simple feature extraction and online classification